Back to EveryPatent.com
United States Patent |
5,262,941
|
Saladin
,   et al.
|
November 16, 1993
|
Expert credit recommendation method and system
Abstract
An expert floorplan credit recommendation method and system, comprising a
database and stored program for arriving at a credit recommendation by
processing data on a computer by means of a decision matrix tree that
emulates the thought processes of credit experts.
Inventors:
|
Saladin; Emery F. (Chesterfield, MO);
Mate; Karol V. (Merrimac, MA);
Gers; Harvey (Chesterfield, MO);
Ruhlin; Kurt A. (Webster Groves, MO)
|
Assignee:
|
ITT Corporation (New York, NY)
|
Appl. No.:
|
502850 |
Filed:
|
March 30, 1990 |
Current U.S. Class: |
705/38 |
Intern'l Class: |
G06F 015/20; G06G 007/52 |
Field of Search: |
364/401,406,408
|
References Cited
U.S. Patent Documents
T998008 | Sep., 1980 | Delano | 364/401.
|
4774664 | Sep., 1988 | Campbell | 364/408.
|
Other References
Burgess, The Electronic Experts, Washington Post, 1 Oct. 1989, H1, H4.
|
Primary Examiner: Envall, Jr.; Roy N.
Assistant Examiner: Bai; Ari M.
Attorney, Agent or Firm: White & Case
Claims
What is claimed is:
1. A computerized expert credit recommendation system useful for credit
analysis automation and training credit analysts, comprising:
a) an input means for entering data in response to screen messages;
b) an output means;
c) a programmable digital computer in association with
(i) a database management computer program, which prompts in a
predetermined sequence of queries which emulate the reasoning of an expert
credit analyst through the output means, a user to enter through the input
means, in a predetermined order, data relating to criteria used in making
credit recommendations, and which manages such input data;
(ii) a knowledge base containing a plurality of predetermined and defined
standardized rating values relating to data on criteria
(iii) an expert credit recommendation computer program; which manipulates
and performs operations on the input data in association with the database
management program and the knowledge base
wherein the digital computer utilizes the expert credit recommendation
computer program in conjunction with the data input by the user to assign
to each of the criteria a standardized rating value from the knowledge
base and to utilize such assigned standardized rating value to simulate a
plurality of decision matrices arranged in a defined decision matrix tree
series, having a financial branch and a dealer rating branch;
wherein, for each branch point in the matrix tree series the system
determines and assigns a defined standardized rating value to a resulting
criterion;
which standardized rating value is applied in a subsequent decision matrix
until a final decision matrix is reached; and
wherein after all the data has been entered by the user in the
predetermined sequence of queries, the system generates an overall rating
which is displayed to the user on the output means specified by the
standardized rating value assigned to a final criterion resulting from the
financial branch and the standardized rating value assigned to a final
criterion resulting from the dealer rating branch.
2. A computerized expert credit recommendation system as in claim 1,
wherein the system further comprises a means for determining a credit
recommendation by comparing the defined standardized rating value assigned
to a net hard collateral criterion against the overall rating so as to
determine and assign a defined standardized rating value to a final rating
criterion, which final rating criterion's assigned defined standardize
rating value is compared against the standardized rating value assigned to
a credit recommendation matrix input criterion to arrive at the credit
recommendation; and a means for communicating the credit recommendation to
the output means.
3. A computerized expert credit recommendation system as in claim 2,
wherein the system further comprises a means for formulating and
communicating to the output means comments on the credit recommendation.
4. A method of determining a credit recommendation, comprising:
a) providing to a computer comprising an input means and an output means, a
database management program and an expert credit recommendation computer
program;
the computer and the expert recommendation computer program operating to
simulate a plurality of decision matrices utilizing standardized rating
values, said matrices being configured in a defined decision matrix tree
series, comprising a financial branch and a dealer rating branch;
wherein the computer and the database management program prompt a user
through the output means, via screen prompts generating in a
pre-determined sequence and communicated to the output means, to enter
data related to specific criteria in a sequence and manner which emulates
the reasoning of an expert credit analyst;
b) entering the data related to the criteria through the input means, in
response to the prompts;
c) determining and assigning a defined standardized rating value to the
data relating to the criteria for which data has been entered, and
applying the standardized rating values assigned to those criteria and to
the resulting criteria to the decision matrices in a predetermined manner
according to the defined decision matrix tree series until a final
decision matrix is reached; and
d) comparing the defined standardized rating value assigned to a final
resulting criterion from the financial branch against a final resulting
criterion from the dealer rating branch in the final decision matrix to
arrive at an overall rating.
5. A method of determining a credit recommendation as in claim 4, further
comprising communicating the overall rating to the output means.
6. A method of determining a credit recommendation as in claim 9, wherein
the overall rating is compared against the standardized rating value
assigned to a net hard collateral criterion so as to determine a
standardized rating value for a final rating criterion; and
comparing the standardized rating value assigned to the final rating
criterion against the standardized rating value assigned to a credit
recommendation matrix input criterion so as to produce the credit
recommendation.
7. A method of determining a credit recommendation as in claim 6, further
comprising communicating the credit recommendation to the output means.
8. A method of determining a credit recommendation as in claim 4, wherein
the financial branch comprises an interim balance sheet decision matrix,
an interim income trend decision matrix, and interim income statement
decision matrix, an interim balance sheet and trend decision matrix, an
interim income statement and trend decision matrix, an interim financial
decision matrix, a balance sheet decision matrix, an income trend decision
matrix, an income statement decision matrix, a balance sheet and trend
decision matrix, an income statement and trend decision matrix, and a
fiscal-year-end decision matrix; and
the dealer rating branch comprises a lender experience decision matrix, an
owner experience decision matrix, a business pay habits decision matrix, a
business experience decision matrix, a business character decision matrix,
a character decision matrix, a manufacturing strength decision matrix, a
dealer collateral decision matrix, a manufacturing collateral decision
matrix and a collateral decision matrix.
9. A computerized expert credit recommendation system as in claim 1,
wherein the system operates to simulate decision matrices configured in a
defined matrix tree series as in FIGS. 1-4.
Description
BACKGROUND AND SUMMARY OF THE INVENTION
Expert systems for making credit recommendations are generally based on
numeric scoring and do not particularly reflect the decision process of a
credit expert. This invention, on the other hand, comprises an expert
credit method and system which use a decision matrix tree to emulate the
decision process of credit experts in analyzing a credit applicant and
recommending whether to extend the credit sought.
The method used is to construct and implement, for example, in a
software-controlled general purpose digital computer, a decision matrix
tree. The tree is constructed so as to emulate the thought processes of
human credit experts. Each matrix in the tree compares two pertinent
characteristics and in turn represents, and provides as output, another
pertinent characteristic which is then available as input to a subsequent
matrix in the tree. In each matrix, each of the two characteristics being
compared is depicted on one axis of the matrix. Each of the two axes of
the matrix has provision for a number of different values of the
characteristic depicted on that axis. Each box in the matrix represents a
different combination of values of those two input characteristics. For
each of these possible combinations, the matrix determines the value of
the characteristic represented by the matrix itself. Thus, each matrix
accepts as input two values, one on each axis. Each matrix delivers as
output a single value, determined by the value of the matrix box defined
by the two values on the axes.
Each input value to a matrix emanates from one of two sources. One or both
input values for a matrix may be output from preceding matrixes in the
matrix decision tree. Other input values are supplied from the "knowledge
base," which is a database resident in or accessible to the computer and
which has been supplied with information derived from credit experts.
The form of the decision matrix tree, as well as the identity of the
characteristics compared in the various matrixes, the output
characteristic from each matrix, the order of the matrices, and the
designations of values in each box of each matrix are all the product of
extensive contribution by credit experts.
A preferred embodiment of the invention employs a general purpose computer
under the control of rule-based expert system artificial intelligence
software that emulates the decision process of experts in the field of
floorplan inventory lending. Industry characteristics and lending
practices unique to floorplan lending play a fundamental role in the
underlying decision matrix tree and characteristics ("criteria") used in
this method and system. As opposed to a numeric scoring method or system,
this invention actually compares various factors used by experts in making
decisions, to emulate more closely the experts' decision process in
reaching a recommendation on whether to extend credit.
In addition to recommending a determination on whether to extend the credit
sought, the method and system of the invention may provide output, usually
in the form of computer print-out or computer screen displays, on matters
supporting the ultimate recommendation. For example, the method and system
may provide as output a summary of the criteria input, and the
recommendation, in a prescribed format. Alternatively or in addition, the
method and system may provide as output a sequence of observations
regarding criteria perceived as positive, and a sequence of observations
regarding criteria perceived as negative. The output may include
observations made directly by the user (and input to the computer) as well
as those generated by the invention. The output from the invention may
also include detailed listings and analyses of various criteria deemed
significant to the recommendation (e.g., information relating to financial
statements). This aspect of the invention is of assistance in making
manifest the rationale behind the recommendation. It is also useful in
training personnel in the credit determination process.
The expert method and system is designed to aid and train floorplan credit
analysts in making credit decisions while ensuring consistency in credit
policy. The system is designed for use on personal computers. The
preferred embodiment operates in connection with IBM.RTM. and
IBM-compatible personal computers having at least a 640K byte memory, a 20
megabyte disk drive and a printer. Analysts input information into the
system, which provides them with results of an extensive analysis both on
screen and through several printed reports.
The method and system involve two interacting components. The first
component is a database management program which also comprises a user
interface (generally a screen display) that prompts an individual analyst
for all information pertinent to a credit decision. In a preferred
embodiment of the invention, PARADOX.RTM. relational data base software
(version 3.01), marketed by Borland International, is used as the database
manager. The information input by the user in response to prompts includes
the type of credit line requested, the level of management experience,
types of collateral and inventory, type of business or industry, credit
history and financial information. Mathematical. calculations are
performed on certain of this information by stored software which
generates financial and other ratios pertinent to a decision. This data is
then maintained in the database for retrieval at any time. The software
may also provide for the electronic transfer of client data to other
locations, e.g., regional offices.
The second component of the method and system is a computer program written
in a language suited to artificial intelligence expert systems, such as
Prolog Version 5.x, marketed by Arity Corporation. This program is the
"brain" of the method and system, analyzing client information that it
obtains from the database manager. The analysis performed by the program
emulates the reasoning patterns of floorplan credit experts by using a
decision matrix tree, comprising a plurality of decision matrices arranged
in branches. To arrive at the final recommendation to grant or deny
credit, or further review the matter, the program must perform each of the
operations in each branch of the decision tree.
At the initial stage, the process and system compare each of a number of
specific criteria to levels or ranges characterized in a particular manner
by credit experts for the specific business or industry. Some of the
relevant criteria are: direct payment experience of the user with the
borrower, payment experience of other creditors, industry experience of
the borrower's management, personal character of management, strength of
industry, strength of product sold, strength of suppliers, type and
strength of collateral, type and strength of documentation, sales level
and trend, gross margins and trend, net margins and trend, leverage
position and trend, and liquidity position and trend.
With the information obtained from this analysis, the process proceeds
through several series of matrices arranged in branches whereby the
strengths and weaknesses of each criterion category are compared with
strengths and weaknesses of one or more other criterion categories. The
result of each such comparison is then used in another comparison in a
subsequent decision. Consideration is given to the significance of each
category relative to the overall decision by its position in the tree.
Those categories deemed more important in the final decision are dealt
with last. This analysis process is unlike credit scoring systems because
numeric weights are not used. Instead, the program actually considers the
strength of one criterion in relation to another criterion, making many
such decisions throughout the analysis process. For example, the program
may compare components of a balance sheet and reach its decision based on
the overall strength of the balance sheet alone. At the same level, the
program is comparing other components to decide the strength of the
balance sheet trend, income statement and the income statement trend. The
analysis continues by then comparing these new decisions to one another.
This method is duplicated for all aspects of the decision process.
As the analysis progresses, and based on the criteria and results of
criteria comparisons, the program constructs comments unique to each
client indicating the status of each important element of floorplan
lending. Upon completion these comments are transferred to the database
management system for storage and retrieval.
Output from the method and system may include an "Executive Summary"
designed to provide a quick look at the condition of the most significant
factors, positive and negative, which pertain to the credit decision.
Comments presented in this summary are generated through the analysis
process and are unique to each client. The output from the method and
system may also include a standard report indicating the result of each
major step in the analysis and a summary of data input. Additionally, the
method and system may provide standardized financial spreadsheets which
include financial ratios relating to inventory financing.
The output from this method and system is used as a tool in making credit
decisions and can also be used for training credit analysts in the proper
requirements of floorplan lending. Use of the method and system helps a
lending organization ensure a uniform standard in the credit evaluation
process and helps to eliminate arbitrary considerations from the decision.
It also provides a means for better and more thorough supervision of the
credit decision processes of subordinates. The system further highlights
key issues, both positive and negative, and by inference indicates those
areas that need to be addressed in order to make a transaction more
favorable.
BRIEF DESCRIPTION OF FIGURES
FIG. 1. Depicts the branch structure of the decision matrix tree comprising
a fiscal year end financial branch, interim financial branch, and a dealer
branch which further comprises a collateral branch and a character branch.
FIGS. 2A and 2B Depicts the financial branch of the decision matrix tree.
FIG. 3. Depicts the collateral branch of the decision matrix tree.
FIG. 4. Depicts the character branch of the decision matrix tree.
DETAILED DESCRIPTION OF THE INVENTION
The specific floorplan expert credit method and system embodiment of the
invention works essentially as follows. A PC-based program written in
Paradox application language guides the credit analyst user through the
data input phase with screen prompts. Necessary calculations may also be
performed by this program on the data. The input data and the results of
those calculations are then stored in the database, which in this case is
the relational database, Paradox. The Paradox application language program
then stores the data and, upon command by the user, formats the required
data in a separate ASCII file. That program then transfers control to the
logic or artificial intelligence expert program, which in this embodiment
is written in Prolog. The logic then retrieves from the ASCII file the
data for analysis, and calls on its knowledge base for the specific
product line-dependent standard data, which is updated for economic or
policy changes periodically, e.g., semi-annually. With the standard data,
the input data and the modified input data the logic then performs its
analysis according to the decision matrix tree depicted in FIGS. 1-4.
After the analysis the logic arrives at a result, i.e., a credit
recommendation, and also generates unique comments according to each
analysis. The results are then processed into a form which is imported
into the database where it is stored and may be displayed or printed.
Referring to FIG. 1, the expert system computer program is comprised of a
decision matrix tree, which is organized into branches. Each branch
comprises one or more decision matrices positioned in sequence according
to the weight or influence on the final decision of the criteria analyzed
in that matrix.
In the invention's preferred embodiment, which concerns floorplan credit
financing, the decision matrix tree is organized into two major branches,
the financial branch FIGS. 2A and 2B and the dealer branch FIGS. 3 and 4.
The financial branch in turn is further differentiated into an interim
financial branch and a fiscal-year-end financial branch, while the dealer
branch is differentiated into a character branch FIG. 4 and a collateral
branch FIG. 3.
In the preferred embodiment each decision matrix in is depicted as a
5.times.5 grid with an "x" and "y" axis. Each axis presents one of the two
criteria compared in that decision matrix. Each axis provides space for
five ratings for the criterion depicted: "good", "satisfactory",
"marginal", "poor" and "unknown". The predetermined values for each rating
for each criterion or factor is stored in the program's knowledge base.
Each box in the grid is designated by one of the same five ratings,
thereby assigning that rating to a situation characterized as having the
criteria ratings specified on the "x" and "y" axes corresponding to that
box. I.e., once the value of the two characteristics being compared is
determined, a box in the grid is defined, and the value of that box is the
value assigned to the output characteristic.
The financial branch of the decision tree, depicted in FIGS. 2A and 2B, is
comprised of decision matrices 1-13. In the preferred embodiment of this
system the financial branch divides into the interim financial sub-branch,
which comprises the decision matrices 1-6, and the fiscal-year-end ("FYE")
sub-branch, comprising matrices 7--12. These two separate sub-branches
have similar form. However, the FYE sub-branch of the decision matrix is
constructed by using data from fiscal year end financial statements,
whereas the interim sub-branch is constructed by using data from interim
financial statements. The operations on the data by the expert program,
however, are the same for both sub-branches. In order to simplify this
description the financial branch operation will be described generically,
with the understanding that the sequence and operations apply both to the
interim and FYE branches.
Balance Sheet decision matrix 1 (for FYE, or 7 for Interim) compares
Liquidity (or Interim Liquidity) for each specific product line with the
Leverage (or Interim Leverage). The term "liquidity" is defined as the
ratio of current assets to current liabilities. The value for the
criterion "Liquidity" is obtained by comparing the specific dealer's
liquidity for the product line in question with standard liquidity ranges
stored in the logic's knowledge base, to obtain a value of "good,"
"satisfactory," "marginal," "poor" or "unknown." The term "leverage" is
defined as debt to tangible net worth. The value of the criterion leverage
is determined by comparing the dealer's leverage with the standard
leverage ranges stored in logic's knowledge base. When values have been
determined for Liquidity and Leverage, the Balance Sheet matrix (which is
stored in the knowledge base) assigns a value to the criterion Balance
Sheet.
The value resulting from the Balance Sheet matrix 1 (or 7) is then further
processed in the Balance Sheet and Trend matrix 2 (or 8), where the
criterion Balance Sheet is compared to the criterion Leverage Trend. The
term "leverage trend" is defined as the change in leverage. Logic obtains
a value for the criterion Leverage Trend by comparing the leverage trend
of this specific dealer with the leverage trend ranges stored in the
knowledge base. The resulting rating of "good", "satisfactory",
"marginal", "poor" or unknown is then compared to the Balance Sheet value
from matrix 1 (or 7). The resulting Balance Sheet Trend value of matrix 2
(or 8) is then used in a comparison in decision matrix 3 (or 9), the
Financial matrix, where it is compared with the value for Income Statement
and Trend, which results from matrix 6 (or 12).
To arrive at a decision from matrix 6 (or 12) the logic program must first
start with processing a decision from matrices 4 and 5 (or 10 and 11). The
Income Trend decision matrix 4 (or 10) consists of a comparison of Sales
Trend, to percent change in Net Profit to Sales ratio. The Sales Trend
criterion is specific to each different product line, and the value is
based on ranges of percentage increases calculated or expected in this
specific product line. These values are stored in the logic's standard
knowledge base. The percent change in the net profit to sales ratio is
calculated based on the dealer's net profit before tax. The specific value
of the criterion Net Profit to Sales for the dealer is assigned based on
the standard ranges stored in the logic's knowledge base. This value is
then compared in matrix 4 (or 10) to Sales Trend to give the value of
Income Trend Criterion. The Income Trend criterion of matrix 4 (or 10) is
then used, as already described, in a comparison in matrix 6 (or 12) with
the Income Statement criterion to determine Income Statement and Trend.
The Income Statement decision matrix 5 (or 11) compares the criterion of
Gross Profit on Sales, to the criterion of Net Profit on Sales. The ratio
of gross profit on sales is calculated for the distinct product line, and
compared to the standard ranges for that product line stored in the
logic's knowledge base, to assign a value to the criterion Gross Profit on
Sales. Once a value is assigned, it is inserted into decision matrix 5 (or
11) and compared to the Net Profit on Sales criterion for the dealer. Net
profit on sales is a percentage ratio calculated for the specific dealer
and compared to the standard ranges for the specific product line stored
in the logic's knowledge base to assign a value to the Net Profit on Sales
criterion. The value of the Income Statement decision matrix 5 (or 11) is
then used in a comparison with the Income Trend criterion from matrix 4
(or 10) in decision matrix 6 (or 12), to determine the Income Statement
and Trend criterion. The criterion derived from the Income Statement and
Trend matrix 6 (or 12) is then used in the Financial decision matrix 3 (or
9), which compares it against the Balance Sheet and Trend criterion in
decision matrix 2 (or 8).
As discussed previously, the decision matrices 1, 2, 3, 4, 5 and 6 for
fiscal year data are in form identical to decision matrices, 7, 8, 9, 10,
11 and 12 for interim data. The values of the Interim Financial matrix 9
and the Fiscal Year End Financial matrix 3 are used to make the comparison
in the Financial matrix 13 to arrive at the Financial criterion. The value
of Financial matrix 13 is then further processed by the logic in
connection with the value determined by the dealer branch, as described
below.
With respect to the Dealer branch, the decision matrix tree is divided into
the Character sub-branch FIG. 4 and the Collateral sub-branch FIG. 3. The
Character sub-branch is comprised of matrices 14, 15, and 21-24. Matrix
21, Owner Experience, compares the criterion of Management Experience at
dealer location to the criterion of Additional Management Experience.
Logic compares the number of years of management experience at the
particular dealer location to the ranges stored in its knowledge base,
assigning the Management Experience criterion a value "good,"
"satisfactory," "marginal," "poor" or "unknown". The same function is also
performed on the Additional Management Experience criterion, which is a
measure of how many additional years of experience current dealer
management has had in a similar industry. Both these values are evaluated
by the logic for this dealer from the standard ranges in the logic
knowledge base, and inserted into the matrix. Logic then evaluates matrix
21, arriving at the criterion of Owner Experience. The resulting value is
then processed in matrix 22, Business Experience.
The Business Experience matrix 22 compare the criterion Company Experience
to the criterion Owner Experience. Company Experience is a function of the
number of years the company has been in business. The specific term for
the dealer in question is compared by the logic to the standard ranges in
the logic's knowledge base, and the value assigned is inserted into matrix
22. After evaluation, the value according the Business Experience matrix
is further processed in matrix 23, Business Character.
In the Business Character matrix the Business Experience decision from
matrix 22 is compared to the Business Pay Habits criterion derived from
matrix 15. To arrive at matrix 15, the logic must first process the
Lender's Experience decision from matrix 14. Matrix 14 compares the Pay As
Sold ("PAS") criterion, which is defined as the percentage of monthly
collections made at the lender's audit of the dealer, to the Schedule Pay
Plan ("SPP") experience of the lender with this dealer, calculated in
terms of average days late on scheduled payment. The PAS and SPP criteria
values for the specific dealer are retrieved by the logic from the
database and compared to the standard ranges stored in the logic's
knowledge base. These values are inserted into the matrix 14, Lender
Experience, and a decision value is calculated.
The Business Pay Habit matrix 15 compares the Lender Experience from matrix
14 to the criterion of Other Creditor Experience. Other Creditor
Experience is established by credit ratings collected by the lender from
other creditors, usually through telephone solicitation by the lender. The
Other Creditor Experience value is determined by the user and input into
the logic's knowledge base. The value is retrieved from the knowledge base
and compared to Lender Experience from matrix 14. The resulting Business
Pay Habit criterion is then further processed in matrix 23 Business
Character, where it is compared to the Business Experience criterion from
matrix 22. In matrix 23, logic calculates a value and this criterion of
Business Character is further processed in the Character matrix 24.
Character matrix 24 compares the criterion of Business Character to the
criterion of Personal Credit Bureau Reports. The Personal Credit Bureau
Reports criterion concerns the personal credit of any guarantors who will
act as guarantors of the loan. Personal Credit Bureau Reports are usually
required when, and only when a personal guaranty is held as additional
collateral on the loan. The value for Personal Credit Bureau Reports is
input into the database and retrieved by the logic. The logic then
calculates a value according to matrix 24 and this criterion is further
processed in the Dealer Rating matrix 20 where it is compared to the value
of the Collateral criterion calculated in the collateral branch of the
matrix FIG. 3, as described below.
The Collateral branch of the matrix comprises matrices 16-19. The
Collateral branch begins with matrix 16, Manufacturer Strength, which
compares the criteria of Manufacturer Buy Back Agreement to Manufacturers'
Financial Rating. These criteria concern the manufacturers of the goods on
the basis of which the floorplan lender is lending to the dealer. The user
determines Manufacturer Financial Rating criteria for each manufacturer of
goods sold, and an overall value for this criterion, and inputs that value
into the database. The value is inserted into matrix 16. The Manufacturer
Buy Back criterion concerns whether the lender has a repurchase agreement
with the manufacturer, and the type of repurchase agreement. The logic
calculates a value for the Manufacturing Strength criterion based on
matrix 16 and this criterion is further processed in matrix 18,
Manufacturing Collateral. In matrix 18, the Manufacturing Strength is
compared to the Product Line criterion. The Product Line criterion is a
function of the percentage of the dealer's credit line to be used on the
manufacturer's products. This percentage, calculated for the individual
dealer and input to the database is compared to the ranges stored in the
logic's knowledge base and a value is inserted into matrix 18.
The resulting decision value is further processed in matrix 19, Collateral.
Matrix 19 compares Manufacturing Collateral with the criterion of Dealer
Collateral, which is derived from decision matrix 17. Matrix 17, Dealer
Collateral, compares the criterion of Soft Collateral with that of
Document Strength. Soft Collateral is defined as the personal tangible net
worth of the guarantor, as a percentage of the credit line. Based on input
to the database, the logic calculates a value for the specific guarantor,
compares this value to the standard ranges in its knowledge base, and then
assigns a value for the dealer, which is inserted into matrix 17. Document
Strength is a measure of the strength of any lien that is placed upon
inventory. The user inputs into the database the nature of the lien, and
the database converts this information into a Document Strength criterion
value.
At this point, both the Financial branch and the Dealer branch have been
completely determined, as a value exists for the Financial criterion in
matrix 13, and a value exists for the Dealer Rating criterion in matrix
20. These two values are input to the Overall Rating criterion in matrix
25, at which point the expert system has arrived at a value that can be
used for making a determination. However, in a preferred embodiment of the
invention further operations are performed in matrices 26 and 27 to give
more weight to certain factors toward a recommendation decision. The value
determined from the Overal Rating matrix 25 is transmitted to the Final
Rating matrix 26, where it is compared with Net Hard Collateral. Net Hard
Collateral is the portion of the credit line covered by letters of credit,
certificates of deposit, and half the value of any mortgages used as
collateral. The dollar amount of hard collateral held on the dealer is
input into the database. The Paradox application language program
calculates Net Hard Collateral as the percent of hard collateral to the
proposed credit line discounted for any "SPP deficit." SPP deficit is
defined as one less than the ratio of outstanding turn (if known to the
lender) or program turn (otherwise), divided by inventory turn. Logic
computes a numerical figure for Net Hard Collateral and compares it to
predetermined stored ranges, whereupon a value is assigned. This value for
Net Hard Collateral is compared with the Overall Rating criterion, to
yield a value for the Final Rating criterion. This value for Final Rating
is then passed on matrix 27, Recommendation.
The Recommendation matrix is unlike any of the other matrices. Although the
"y" axis (Final Rating) has the same five values as do the other matrices,
the "x" axis has the following values: credit line is more than 300% of
average inventory; net SPP deficit is greater than 10%; current ownership
exceeds ten years; dealer has been with lender for more than five years;
and Otherwise. Further, the values in each of the 25 boxes comprising the
matrix are "Recommend," "Review," and "Not Recommend." This is the final
matrix and it embodies the ultimate result of the system. The
recommendation is then displayed on screen or by some other output means
to the operator who uses it in making his or her credit decision. Further,
supporting output may be printed, e.g., in the form of a summary of input
data, a sequence of observations generated by the method and system
regarding positive and negative criteria, financial data, or an Executive
Summary.
The criteria analyzed by the system and compared in the matrices depicted
in FIGS. 1, 2, 3 and 4 are listed in Table 1.
TABLE 1
______________________________________
PRODUCT MAR-
CRITERIA LINE GOOD SAT GINAL
______________________________________
pas experience
-- 35 50 55
spp experience
-- 3 5 7
company -- 5 3 1
experience
additional mgt
-- 10 5 3
experience
management -- 4 2 0.5
experience
personal -- 1 2 3
character
other creditor
-- 1 2 3
experience
product line
-- 75 50 40
document strength
-- 1 2 3
manufacturer
-- 1 2 3
buyback
manufacturer
-- 1 2 3
financial rating
hard collateral
-- 70.0 35.0 10.0
soft collateral
-- 50.0 0.0 -50.0
gross profit to
mis 25 20 17
sales
gross profit to
tv 24 20 18
sales
gross profit to
marine 23 18 15
sales
gross profit to
mcycle 20 18 16
sales
gross profit to
rv 19 15 13
sales
gross profit to
mob home 18 15 13
sales
gross profit to
ind eq 23 20 18
sales
gross profit to
office 24 20 17
sales
gross profit to
keyboard 33 27 22
sales
gross profit to
unknown 25 20 18
sales
interim gp sa
mis 25 20 17
interim gp sa
tv 24 20 18
interim gp sa
marine 23 18 15
interim gp sa
mcycle 20 18 16
interim gp sa
rv 19 15 13
interim gp sa
mob home 18 15 13
interim gp sa
ind eq 23 20 18
interim gp sa
office 24 20 17
interim gp sa
keyboard 33 27 22
interim gp sa
unknown 25 20 18
net profit to sales
mis 2.0 0.8 -0.2
net profit to sales
marine 2.0 0.8 -0.2
net profit to sales
tv 1.5 0.8 -0.2
net profit to sales
mcycle 1.4 1.0 -0.2
net profit to sales
rv 1.0 0.5 -0.2
net profit to sales
mob home 1.0 0.5 -0.2
net profit to sales
ind eq 2.2 1.0 -0.2
net profit to sales
office 1.5 0.8 -0.2
net profit to sales
keyboard 2.1 1.5 -0.2
net profit to sales
unknown 1.5 0.5 -0.2
interim np sa
mis 2.0 0.8 -0.2
interim np sa
marine 2.0 0.5 -0.2
interim np sa
tv 1.5 0.8 -0.2
interim np sa
mcycle 1.4 1.0 -0.2
interim np sa
rv 1.0 0.5 -0.2
interim np sa
mob home 1.0 0.5 -0.2
interim np sa
ind eq 2.2 1.0 -0.2
interim np sa
office 1.5 0.8 -0.2
interim np sa
keyboard 2.1 1.5 -0.2
interim np sa
unknown 1.5 0.5 -0.2
sales trend
mis 10 5 0
sales trend
marine 5 0 -5
sales trend
mcycle 0 -5 -7
sales trend
tv 5 0 -5
sales trend
rv 2 -3 -5
sales trend
mob home 4 0 -5
sales trend
ind eq 4 0 -5
sales trend
office 8 4 0
sales trend
keyboard 3 0 -3
sales trend
unknown 5 0 -5
interim sales trend
mis 10 5 0
interim sales trend
marine 5 0 -5
interim sales trend
mcycle 0 -5 -7
interim sales trend
tv 5 2 -3
interim sales trend
rv 2 -3 -5
interim sales trend
mob home 4 0 -5
interim sales trend
ind eq 4 0 -5
interim sales trend
office 8 4 0
interim sales trend
keyboard 3 0 -3
interim sales trend
unknown 5 0 -5
net profit to sales
-- 0.3 -0.3 -0.5
trend
interim mp sa
-- 0.3 -0.3 -0.5
trend
leverage trend
-- 1.2 1.5 2.0
interim leverage
-- 1.2 1.5 2.0
trend
leverage -- 4.0 6.0 12.0
interim leverage
-- 4.0 6.0 12.0
liquidity mis 1.4 1.1 1.0
liquidity marine 1.3 1.1 1.0
liquidity mcycle 1.2 1.1 1.0
liquidity tv 1.2 1.1 1.0
liquidity rv 1.3 1.1 1.0
liquidity mob home 1.2 1.0 0.9
liquidity ind eq 1.2 1.0 0.9
liquidity office 1.4 1.1 1.0
liquidity keyboard 1.3 1.0 0.9
liquidity unknown 1.3 1.1 1.0
interim liquidity
mis 1.4 1.1 1.0
interim liquidity
marine 1.3 1.1 1.0
interim liquidity
mcycle 1.2 1.1 1.0
interim liquidity
tv 1.2 1.1 1.0
interim liquidity
rv 1.3 1.1 1.0
interim liquidity
mob home 1.2 1.0 0.9
interim liquidity
ind eq 1.2 1.0 0.9
interim liquidity
office 1.4 1.1 0.9
interim liquidity
keyboard 1.3 1.0 0.9
interim liquidity
unknown 1.3 1.1 1.0
______________________________________
Top